Japanese Interest Rate Swap Spreads and Monetary Policy in the Framework of VAR ☆ Takayasu Ito Summary This paper investigates the determinants of 2-year, 5-year and 10-year interest rate swap spreads in Japan by considering the difference of monetary policy regimes by the Bank of Japan (BOJ) in the Framework of VAR (Vector Auto Regression). Four determinants of swap spreads – default risk, liquidity premium, the slope of yield curve and volatility - are chosen. Swap spread’s own shocks play a dominant role in the determination of swap spreads in each sample. The slope of yield curve plays an important role for 10- year swap spreads in the period of loosening monetary policy. In the tightening cycle, volatility is a contributor for 10 year swap spread. The default risk is a major contributor only for 5-year swap spread in easing period. On other maturities and samples, default risk is not a determinant of swap spread. Keywords: Swap Spread, VAR, Monetary Policy JEL Classifications: E43, E52 ☆ The research for this paper is supported by Grant-in Aid for Scientific Research (KAKENHI 9530271) from JSPS. 1
21
Embed
Japanese Interest Rate Swap Spreads and Monetary …Swap spread’s own shocks play a dominant role in the determination of swap spreads in each sample. The slope of yield curve plays
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Japanese Interest Rate Swap Spreads and Monetary Policy
in the Framework of VAR ☆
Takayasu Ito
Summary
This paper investigates the determinants of 2-year, 5-year and 10-year interest rate swap
spreads in Japan by considering the difference of monetary policy regimes by the Bank of
Japan (BOJ) in the Framework of VAR (Vector Auto Regression). Four determinants of
swap spreads – default risk, liquidity premium, the slope of yield curve and volatility -
are chosen. Swap spread’s own shocks play a dominant role in the determination of swap
spreads in each sample. The slope of yield curve plays an important role for 10- year
swap spreads in the period of loosening monetary policy. In the tightening cycle,
volatility is a contributor for 10 year swap spread. The default risk is a major contributor
only for 5-year swap spread in easing period. On other maturities and samples, default
risk is not a determinant of swap spread.
Keywords: Swap Spread, VAR, Monetary Policy
JEL Classifications: E43, E52
☆ The research for this paper is supported by Grant-in Aid for Scientific Research (KAKENHI 9530271)
from JSPS.
1
VAR を用いた日本円金利スワップスプレッドと金融政策の分析
要旨
本稿では日本銀行の金融政策スタンスを考慮に入れて、日本円金利スワップスプレ
ッド(2 年物、5 年物、10 年物)の決定要因を VAR (Vector Auto Regression)を用いて分
1.Introduction This paper investigates the determinants of Japanese interest rate swap spreads in each
sub sample by considering monetary policy regimes. Four determinants of swap spreads -
corporate bond spread, TED spread, the slope of yield curve and volatility - are chosen.
The monetary policy changes tend to exert an impact on financial markets. The
asymmetric impacts of monetary policy on interest swap spreads are focused by dividing
the whole sample period into three depending on monetary policy regime.
First period is from February 15, 1999 through August 11, 2000. The Bank of Japan
(BOJ) adopted the zero interest rate policy to counter deflationary pressure. Second
period is from March 21, 2001 through March 9, 2006. The BOJ introduced the
quantitative easing policy under deflation caused by bad loan problem and weak domestic
demand. Third period is from March 10, 2006 through July 31, 2008.After the BOJ lifted
the quantitative easing policy, they hiked uncollateralized call rate twice. In terms of
monetary policy regimes, first period and second period are easing, third period is
tightening.
An interest rate swap is an agreement between two parties to exchange cash flows in the
future. In a typical agreement, two counterparties exchange streams of fixed and floating
interest rate payments. Thus fixed interest rate payment can be transformed into floating
payment and vice versa. The amount of each floating rate payment is based on a variable
rate that has been mutually agreed upon by both counterparties. For example, the floating
rate payment could be based on 6 month LIBOR (London Interbank Offered Rate).
The market for interest rate swaps has grown exponentially in the 1990’s. According to a
survey by BIS (Bank for International Settlements), the notional outstanding volume of
transactions of Japanese yen interest rate derivatives amounted to 58,056 billions of US
dollars at the end of June 2008 1 . Differences between swap rates and government bond
yields of the same maturity are referred to as swap spreads. If the swap and government
bond markets are efficiently priced, swap spreads may reveal something about the
perception of the systemic risk in the banking sector.
As for the analysis of the interest rate swap spreads in US market, previous studies such
as Sun et al (1993), Brown et al (1994), Duffie and Huang (1996), Cossin and Pirotte
(1997), Minton (1997),Lang et al (1998), Lekkos and Milas (2001), Fehle (2003), and
Huang and Chen (2007) are cited.
1 Statistics are cited from OTC Derivatives Market Activity in the first half of 2008. At the end of June
1998, the notional outstanding volume of transactions of yen interest rate derivatives was 7,164 billions of US dollars. For details, see BIS (1998) and BIS (2008).
3
Sun et al (1993) examine the effect of dealers' credit reputations on swap quotations and
bid-offer spreads by using quotations from two interest rate swap dealers with different
credit ratings (AAA and A). The AAA offer rates are significantly higher than the A offer
rates, and the AAA bid rates are significantly lower than the A bid rates. They also
document the relation between swap rates and par bond yields estimated from London
interbank offered rate (LIBOR) and bid rate (LIBID) data. They identify some of the
problems in testing the implications of swap pricing theory.
Duffie and Huang (1996) present a model for valuing claims subject to default by both
contracting parties, such as swaps and forwards. With counterparties of different default
risk, the promised cash flows of a swap are discounted by a switching discount rate that,
at any given state and time, is equal to the discount rate of the counterparty for whom the
swap is currently out of the money (that is, a liability). The impact of credit-risk
asymmetry and of netting is presented through both theory and numerical examples,
which include interest rate and currency swaps.
Brown et al (1994) analyze US swap spreads to find that 1) short- term, 1-, and 3-year
swaps are priced differently from longer-term, 5-, 7-, and 10-year swaps; and 2) the
pricing dynamics for all five swap maturities changed substantially during the period
spanning January 1985 to May 1991. Cossin and Pirotte (1997) conduct empirical
analysis on transaction data and show support for the presence of credit risk in swap
spreads. Credit ratings appear to be a significant factor affecting swap spreads not only
for their pooled sample but for IRS and for CS separately as well. In IRS, the credit rating
impact on prices seems to come largely at the detriment of the non-rated companies.
Lang et al (1998) argue that an interest rate swap, as a non-redundant security, creates
surplus which will be shared by swap counterparties to compensate their risks in swaps.
Analyzing the time series impacts of the changes of risks of swap counterparties on swap
spreads, they conclude that both lower and higher rating bond spreads have positive
impacts on swap spreads.
Lekkos and Milas (2001) assess the ability of the factors proposed in previous research
to account for the stochastic evolution of the term structure of the U.S. and U.K. swap
spreads. Using as factor proxies the level, volatility, and slope of the zerocoupon
government yield curve as well as the Treasury-bill–London Interbank Offer Rate
(LIBOR) spread and the corporate bond spread, they identify a procyclical behavior for
the short-maturity U.S. swap spreads and
a countercyclical behavior for longer maturity U.S. swap spreads. Liquidity and
corporate bond spreads are also significant, but their importance varies with maturity.
4
Minton (1997) directly tests the analogy between short-term swaps and Eurodollar strips
and finds that fair-value short-term swap rates exist in the Eurodollar future market.
However, proxies for differential probability of counterparty default are statistically
significant determinants of the difference between OTC swap rates and swap rates
derived from Eurodollar futures prices for maturities of three and four years.
Fehle (2003) analyzes 2- year and 5-year swap spreads in 7 countries (US, UK, Japan,
Germany, France, Spain and Netherland). They conclude that corporate bond spread,
LIBOR spread and slope of the yield curve are components of swap spreads.
Huang and Chen (2007) analyze the asymmetric impacts of various economic shocks on
swap spreads under distinct Fed monetary policy regimes. The results indicate that (a)
during periods of aggressive interest rate reductions, slope of the Treasury term structure
accounts for a sizeable share of the swap spread variance although default shock is also a
major player. (b) On the other hand, liquidity premium is the only contributor to the
2-year swap spread variance in monetary tightening cycles. (c) The impact of default risk
varies across both monetary cycles and swap maturities. (d) The effect of interest rate
volatility is generally more evident in loosening monetary regimes.
On the other hand, the number of previous studies analyzing the market other than US is
small. Castagnetti (2004) analyzes the interest rate swap spreads in Germany. Hmano (1997),
Eom et al (2000), Ito (2007) focus on the swap spreads in Japanese market. Hamano
(1997) focuses not on credit risk but on market factors such as TED spread and finds that
swap spreads reflect TED spread and longer term swap spreads are less influenced by
TED spread. On the other hand, Eom et al (2000) focuses on the credit risk and concludes
that yen swap spread is significantly related to proxies for the long term credit risk factor.
Ito (2007) investigates the determinants of interest rate swap spreads in Japan. Four
determinants of swap spreads - TED spread, corporate bond spread, interest rate and the
slope of yield curve from July 12, 1995 through January 31, 2005- are chosen. The swap
spreads of 2 years through 4 years are mostly influenced by TED spread, interest rate and
slope. The swap spread of 5 years is mostly decided by corporate bond spread and slope.
The swap spreads of 7 years and 10 years are mostly affected by corporate bond spread.
Ito (2009) conducts the analysis of Japanese interest rate swap spreads in different
monetary policy regimes by OLS. When monetary policy was tightening, 10 year swap
spread decreased in accordance with the increase of corporate bond spread. TED spread
contributed to swap spreads positively in all maturities under tightening cycle of the
monetary policy. Slope of yield curve contributed more actively to the swap spreads in all
maturities in quantitative easing period and to the swap spreads of 5 years, 7 years and 10
5
years in tightening aspect.
Huang and Chen (2007) is the previous study taking into consideration monetary policy
regimes. This paper will be the first one to analyze interest rate swap spreads in Japan in
different monetary policy regimes by time series model like VAR. Huang and Chen
(2007) analyzes swap spreads of 2 years and 10 years. This paper adds 5 year swap spread
for the analysis. This is because Japanese financial institutions actively participate in the
5 year zone of interest rate swap market.
2.Determinants of Swap Spread 2.1 Default Risk
According to Minton (1997), Brown et al (1994), Eom et al (2000), Lekkos and Milas
(2001), the default risk in swaps can be proxied with the information from the corporate
bond market. Any such proxy is imperfect as mentioned in the previous studies because
the characteristics of the swap and corporate bond are not totally comparable.
Nevertheless, since swap default spreads are unobservable, the difference between the
yield on a portfolio of corporate bonds and the yield on an equivalent government bond
can be used as a proxy for the default premium.
2.2Liquidity Premium
For instance, during periods of weak economy, treasury bonds are considered more
liquid, and swaps thus command a larger liquidity premium. Liquidity effect may be
absent in the aggregate data, but can be arguably pronounced under certain market
conditions. Hamano (1997),Minton (1997), Brown et al (1994), Eom et al (2000), Lekkos
and Milas (2001) check the influence of TED (LIBOR – T-bill ) spread .
Hamano (1997) finds that Japanese yen swap spreads are influenced by TED and their
influences get weaker as the maturities of spread get longer from 1992 through 1996. On
the other hand Eom et al (2000) find that the influences of TED on Japanese swap spreads
get stronger as the maturities of spread get longer from 1990 through 1996.
2.3 Slope of Yield Curve and Volatility
Following the Sorensen and Bollier (1994) framework, in which the slope of the term
structure and interest rate volatility determine the value of the option to default, we
incorporate these two variables into empirical model. It is notable that the impacts of the
yield curve and interest rate volatility on swap spreads may not be symmetrical under
various market conditions. For example, due to investors' risk aversion, risk premium
6
may not necessarily be as responsive to the changes in interest rate volatility during
periods of little default risk.
Similarly, as Huang and Chen (2007) describes, swap spreads may be more responsive
to the shape of yield curve during periods of a steep yield curve due to the “flight to
quality” concern. Aggregating time series data over different market conditions, therefore,
produces results that are in favor of finding no impact of economic shocks on swap
spreads because asymmetrical impacts may cancel out over monetary policy cycles. Eom
et al (2000) find that swap spreads are negatively related to the slope of the term structure.
Huang and Chen (2007) use slope of yield curve and volatility. They calculate volatility
of 2-year US Treasury note by using EGARCH model.
3.Data A total of more than nine years of daily data ranging from February 15, 1999 through
July 31, 2008 are chosen. These data are provided by the Mitsubishi UFJ Securities.
Three monetary policy regimes by the BOJ are chosen. The whole sample from February
15, 2000 through July 31, 2008 is divided into three depending upon the monetary policy
regimes by the BOJ. First period (Sample A) is from February 15,1999 through August 11,
2000. The BOJ continued the zero interest rate policy during first period. Second period
(Sample B) is from March 21, 2001 through March 9, 2006. The BOJ continued the
quantitative easing policy during second period.
Third period (Sample C) is from March 10, 2006 through July 31, 2008. The monetary
policy was tightening after the BOJ announced the lifting of quantitative easing policy on
March 9, 2006. Afterwards the BOJ hiked uncollateralized call rate twice 2 . The period
from August 14, 2000 through March 19, 2001 was excluded from the analysis since the
daily sample is not long enough for the analysis. In terms of monetary policy, the first and
second periods are easing, but the third period is tightening. Variables for the analysis are
defined as below.
3.1 Japanese Interest Rate Swap Spread
Japanese interest rate swap rate minus Japanese government bond yield in the
corresponding maturity is defined as swap spread. S2 is 2- year swap spread. S5 is 5- year
swap spread. S10 is 10- year swap spread.
As for Japanese government bond yield, par rates of Japanese Government Bond (JGB)
2 The BOJ hiked the target of uncollateralized call rate from 0 % to 0.25 on July 14,2006 and from 0.25 %
to 0.5% on February 21,2007.
7
are used 3 . These par rates for the maturities of 2 years, 5 years and 10 years are
calculated by the method mentioned in Adams and Van Deventer (1994) on a daily basis
from February 15, 1999 through July 31, 2008. These par rates are provided by the
Mitsubishi UFJ Securities.
As for the Japanese interest rate swap market, rates of 2 years, 5 years and 10 years as of
3 pm at Tokyo time are used on a daily basis from February 15, 1999 through July 31,
2008. Bid rates and offer rates are indicated in the market. The swap rates provided by the
Mitsubishi UFJ Securities are averages of bid rates and offer rates. The movements of
swap spreads in 2 years, 5 years, and 10 years in each sample period are shown in Figure
1. The descriptive statistics of swap spreads in whole sample and in each sample period
are provided in Table 1.
In Sample B swap spreads of 5 years and 10 years sometimes became negative. In other
words, swap rates sometimes were lower than JGB yields. This phenomenon is peculiar to
Japanese interest rate swap market. Japanese banks activated receiving in Japanese Yen
Interest Rate Swap market to increase profit 4 in especially from 2001 through 2002. Thus
Japanese Interest Rate Swap rates decreased sharply and swap spreads in 5 years and 10
years became negative.
The credit rating of Japanese Government Bond was downgraded a couple times from
2001 through 2003. This increased the risk premium of Japanese Government Bond. Thus
the decrease of yield in Japanese Government was smaller in comparison with interest
rate swap rate even though the BOJ adopted unprecedented easing policy.
3 JGBs are traded on a simple yield. Par rates are compounded yield. 4 The extension of abolishing macro hedge accounting for another year promoted receiving activity. It
was abolished on March 31, 2003.
8
Figure 1 Swap Spreads
% Sample A = from February15, 1999 through August 11,2000.
% Sample B = from March 22, 2001 through March 9, 2006.
% Sample C = from March 10, 2006 through July 31, 2008.
0.00
0.10
0.20
0.30
0.40
0.50
0.60
03/10/06
04/18/06
05/29/06
07/04/06
08/10/06
09/15/06
10/25/06
12/04/06
01/15/07
02/21/07
03/30/07
05/10/07
06/15/07
07/24/07
08/29/07
10/09/07
11/14/07
12/21/07
02/05/08
03/13/08
04/21/08
05/30/08
07/07/08
S2
S5
S10
0.00
0.10
0.20
0.30
0.40
0.50
0.60 02
/15/99
03/15/99
04/15/99
05/15/99
06/15/99
07/15/99
08/15/99
09/15/99
10/15/99
11/15/99
12/15/99
01/15/00
02/15/00
03/15/00
04/15/00
05/15/00
06/15/00
07/15/00
S2
S5
S10
‐0.15
‐0.10
‐0.05
0.00
0.05
0.10
0.15
0.20
0.25
03/21/01
06/21/01
09/21/01
12/21/01
03/21/02
06/21/02
09/21/02
12/21/02
03/21/03
06/21/03
09/21/03
12/21/03
03/21/04
06/21/04
09/21/04
12/21/04
03/21/05
06/21/05
09/21/05
12/21/05
S2
S5
S10
9
Table 1 Descriptive Statistics of Swap Spreads
Variable Average SD Min Max Median
Sample A n=369
S2 0.145 0.043 0.023 0.273 0.148
S5 0.199 0.055 0.048 0.328 0.186
S10 0.363 0.078 0.209 0.554 0.336
Sample B n=1222
S2 0.080 0.022 0.023 0.148 0.081
S5 0.057 0.052 -0.086 0.191 0.067
S10 0.019 0.059 -0.101 0.221 0.016
Sample C n=591
S2 0.207 0.091 0.077 0.521 0.160
S5 0.188 0.049 0.115 0.388 0.180
S10 0.189 0.039 0.117 0.305 0.184
Sample A = from February15, 1999 through August 11,2000.Sample B = from March 21, 2001 through March 9, 2006.Sample C = from March 10, 2006 through July 31, 2008.S2= 2 - year swap spread S5= 5 - year swap spreadS10 = 10 - year swap spread
3.2Determinants of Swap Spread
Default Risk
Default risk is defined as yield spread between 10- year corporate bond issued by the
Tokyo Electric Power Company and 10 year JGB par yield. Corporate bond spread is
considered to represent credit risk. In Japan corporate bond market is illiquid. Thus 10-
year corporate bond issued by the Tokyo Electric Company is the only data available for
the analysis. As for the data source the period from February 15,1999 through July 23,
2007 is from Mitsubishi UFJ Securities. The period from July 24, 2007 through July 31,
2008 is provided by Japan Securities Dealers Association (JASDA).
Liquidity Premium
Liquidity premium is defined as TED spread between 6- month TIBOR and 6 month TB
(Treasury Bill). TIBOR is provided by the Japanese Bankers Association. TB yields are
provided by the Mitsubishi UFJ Securities.
Slope of Yield Curve
Slope of yield curve is defined as the differential between 2 - year and 10 - year JGB par
10
yields as in Huang and Chen (2007). These par rates are provided by the Mitsubishi UFJ
Securities.
Volatility
Yield volatility calculated by EGARCH model is defined as volatility 5 . The 2- year JGB
par rates provided by the Mitsubishi UFJ Securities are used for the calculation since
Huang and Chen (2007) use the 2 year US Treasury note for the calculation of EGARCH
volatility.
The descriptive statistics of determinants of swap spreads in each sample period are
provided in Table 2.
Table 2 Descriptive Statistics of Determinats of Swap Spreads
Sample A = from February15, 1999 through August 11,2000.Sample B = from March 21, 2001 through March 9, 2006.Sample C = from March 10, 2006 through July 31, 2008.CBS=default risk,TED=Liquidity Premium, SLOPE=slope of yield curve VOLA=volatility
5 See Nelson (1991) as for EGARCH model.
11
4.Framework of Analysis Through the analysis of VAR (Vector Auto Regression) model, it’s possible to assess the
impact of economic factors on swap spreads. As Huang and Chen (2007) describes, the
model allows the exogenous variables to have lagged effects on swap spreads, such that
many time series data properties are better described. In doing so, instead of estimating
each regression equation independently, the VAR permits individual regressions to be
estimated within a system where all economic variables are endogenously determined.
This point is also important for this analysis of Japanese interest rate swap spreads
because many variables are closely and endogenously related to each other. For example,
the slope of the yield curve is used as a measurement of economic outlook, but it could
also be argued that the slope is a proxy of default risk.
VAR system is written as follows:
Here is a vector of swap spreads (S2 or S5 or S10), default risk, liquidity premium, slope of yield curve and interest rate volatility. is a 5 x 5 matrix of polynomials in the lag operator L. is a vector of the corresponding innovation processes with zero mean and a non-diagonal variance-covariance matrix of
. The system contains five equations and each one of the five endogenous variables depends on its own lagged values as well as the lagged values of the other four. The optimal lag length is decided by the AIC (Akaike Information Criterion). VAR is supposed to handle stationary data. ALL the data except for interest rate volatility are found to be non- stationary. Thus the first difference of S2, S5, S10, default risk, liquidity premium, slope of yield curve and level data of interest rate volatility are used for the estimation of VAR. After the estimation of VAR, the impacts of economic shocks on swap spreads are investigated by block exogeneity Wald test, variance decompositions and impulse response functions.
5.Result 5.1 Block Exogeneity Wald Test
Table 3 reports whether determinants of swap spreads Granger-cause swap spreads. In
Sample A, the causality from liquidity premium on 5 year swap spread and causalities of
slope and volatility on 10 year swap spread are confirmed. In Sample B, the causalities of
12
default risk and slope on 5 year and the causalities of liquidity and slope on 10 year swap
spread are found. In Sample C, the causality of slope on 5 year swap spread and the
causality of volatility on 10 year swap spread are confirmed. In all samples, none of the
determinants shows predictive power on 2 year swap spread. The lag lengths decided by
AIC in all cases are two. Thus the analysis of prolonged effects of economics shocks on
swap spreads are rather limited when the fact is considered that data used in this paper are
daily based.
Table 3 Bolck Exogeneity Test
Granger Causality Chi-sq Prob Chi-sq Prob Chi-sq ProbSample A Sample B SampleC